Improved Neural Machine Translation with a Syntax-Aware Encoder and Decoder
2017pp. 1936–1945
Citations Over TimeTop 1% of 2017 papers
Abstract
Most neural machine translation (NMT) models are based on the sequential encoder-decoder framework, which makes no use of syntactic information. In this paper, we improve this model by explicitly incorporating source-side syntactic trees. More specifically, we propose (1) a bidirectional tree encoder which learns both sequential and tree structured representations; (2) a tree-coverage model that lets the attention depend on the source-side syntax. Experiments on Chinese-English translation demonstrate that our proposed models outperform the sequential attentional model as well as a stronger baseline with a bottom-up tree encoder and word coverage.
Related Papers
- → Homologous detection based on text, Token and abstract syntax tree comparison(2010)4 cited
- → A Program Similarity Evaluation using Keyword Extraction on Abstract Syntax Tree(2005)2 cited
- → Using circular programs for higher-order syntax(2013)2 cited
- → Syntax Analysis: The Left-Most-Derivation-and-Reduction Trees and its Compare with the LR Parsing Methods(2014)
- BNF syntax unifying Dmoz and Yahoo syntax and its implementation(2009)